I'm creating a bootable Linux system on a PicoZed board (ARM CortexA9 core), and I've run into a "limitation", which I don't think really is a limitation (I get the feeling it's another problem masquerading as a limitation).
I boot by starting the system in JTAG boot mode; after powering on the board I use the xmd debugger to place u-boot into the system's RAM and then I run it.
Next I place the kernel (uImage), the gzip'd initramfs image and the device tree into memory. Finally I tell u-boot to boot the system using bootm, and using three arguments to point out the memory locations of all the images.
All of this works, and I manage to boot up Linux + userland. However, I need to grow the initramfs, and this is where I run into problems.
The working image is 16MiB exactly. I tried to make it 24MiB (completely regenerated from scratch each time I try to boot), but just after the kernel has loaded and it tries to find init the kernel reports file system faults and fails. There shouldn't be any overlaps, but just in case I tried moving things around a little, but the same problem occurred.
After searching for some tips, I saw someone on a forum say that the image needs to be placed at a 16MiB alignment (which I don't think is true, but I tried it none the less, but it didn't get a working system). Another post claimed that the images must be aligned with their sizes (which I again don't think is true, but I tried that as well, but with no change). Yet another post claimed that this happens if the initramfs image crosses the __init end boundary, and that placing the initramfs image firmly inside will allow the memory to be reclaimed after the image has been uncompressed by the kernel, and placing it beyond the __init section will work but then that memory is forever lost after boot. I know way too little about Linux in order to know if any of this is in any way true/accurate, and I have no idea where "__init"'s - if such a thing exists - end-boundary is, but if the issue is that I was crossing it I tried moving the initramfs image way beyond anywhere were I was previously using it, but that didn't change anything.
I also tried the original location (which works with a 16MiB image) and create a 16MiB + 1K sized image, but this didn't work either. (Checking that it didn't overlay any of the over images, obviously).
This originally led me to think there's a 16MiB initramfs size limit lurking somewhere. But on searching for it, it got me thinking that doesn't make sense -- as far as I can gather the bootm command in u-boot shoud set up the tag list for the system (which includes the location and size of the initramfs), and I haven't come across any note about a 16MiB limit in relation to the tagged lists for the initramfs so far.
I have found a page which claims that the initramfs size is in practical terms limited to roughly half the size of physical ram in the system, and the PicoZed board has 1G RAM, so we're in orders of magnitude away from what "should" be a limitation.
To clarify: The 16MiB image is the size of the raw image; compressed it's just under 6MiB. However, if I populate it so it's not compressed as tightly it doesn't make any difference -- the problem doesn't appear to be related to the size of the compressed image, only the uncompressed image.
Main question:
Where is this apparent 16MiB initramfs size limit coming from?
Side-question:
Is there such a thing as an "kernel __init section" which is reclaimed by the kernel after it has uncompressed/loaded images? If so, how do I see/configure the location/size of it?
Size constraints of initramfs on ARM?
I have not encountered a 16 MiB size constraint as you allege.
At one time I thought there was a size limit too, but that turned out to be a memory footprint issue during boot. Once that was sorted out I've been using large initramfs of 30MiB (e.g. with glibc, gtstreamer and qt5 libraries).
Where is this apparent 16MiB initramfs size limit coming from?
There isn't one. A ramfs is only constrained by available RAM.
There is a definition for the "Default RAM disk size", but this would not affect the size of an initramfs.
Your method of booting with the U-Boot bootm command is suspect, i.e. you're passing the memory address of the initramfs as the second argument.
The U-Boot documentation clearly describes the second argument as "the address of an initrd image" (emphasis added).
There is no mention of initramfs as an argument.
Linux kernel documentation states that the initramfs archive can be "linked into the linux kernel image". There's a kernel menuconfig entry for specifying the path to the initramfs cpio file. The make script will append this cpio file to the kernel image so that there is a single image file for booting.
Or (like an initrd) "a separate file" can be passed to the kernel at boot to populate the initramfs.
U-Boot passes the location (and length) of this archive either as an ATAG entry or as a reserved memory region in the Device Tree blob.
The kernel expects either a cpio archive for the initramfs or a tar archive for an initrd.
You neglect to mention (besides its compression) what kind of archive this "initramfs" or "separate file" actually is .
So it's not clear if you're booting the kernel with an initrd (tar archive) or an initramfs (cpio archive).
Your repeated reference to the initramfs as an "image" file rather than a cpio archive suggests that you really are using an initrd.
An initrd would definitely have a size constraint.
Is there such a thing as an "kernel __init section" which is reclaimed by the kernel after it has uncompressed/loaded images?
Yes, there is an __init section of memory, which is released after kernel initialization is complete.
If so, how do I see/configure the location/size of it?
Usually routines and data that have no use after initialization can be declared with the __init macro. See this.
The location and size of this memory section would be under the control of the linker script, rather than explicit user control. The kernel System.map file should have the info for review.
I think you need to pass ramdisk_size in uboot bootargs command
ramdisk_size needs to be set if the ram-disk uncompress file
size is bigger than default setting. It should be more than
ramdisk uncompress file size.
Related
While building the kernel I am giving LOADADDR as "0x80008000":
make uImage LOADADDR=0x80008000
Can you please help to understand what is the use of this? Can I change the LOADADDR, is there any restriction on the length of the LOADADDR?
(I'm assuming that you're using ARM based on the mention of U-Boot and the value of LOADADDR.)
Can you please help to understand what is the use of this?
LOADADDR specifies the address where the kernel image will be located by the linker. (This is true for a few architectures (e.g. Blackfin), but not for ARM.
LOADADDR specifies the address where the kernel image will be located by U-Boot and is stored in the U-Boot header by the mkimage utility. Typically the load address (for placement in memory) is also the start address (for execution). Note that the uImage file is typically just the (self-extracting, compressed) zImage file with the U-Boot wrapper.
Can I change the LOADADDR,
Yes, but according to (Vincent Sanders') Booting ARM Linux that would be contrary to ARM convention:
Despite the ability to place zImage anywhere within memory,
convention has it that it is loaded at the base of physical RAM plus
an offset of 0x8000 (32K). This leaves space for the parameter block
usually placed at offset 0x100, zero page exception vectors and page
tables. This convention is very common.
(The uImage mentioned in your question is probably just a zImage with the U-Boot wrapper, so the quotation does apply.)
is there any restriction on the length of the LOADADDR?
The "length"? If you're using a 32-bit processor, then the length of this address would be 32 bits.
ADDENDUM
arch/arm/boot/Makefile only uses LOADADDR for building the uImage from the zImage.
From (Russel King's) Booting ARM Linux the constraints on this LOADADDR are:
The
kernel should be placed in the first 128MiB of RAM. It is recommended
that it is loaded above 32MiB in order to avoid the need to relocate
prior to decompression, which will make the boot process slightly
faster.
When booting a raw (non-zImage) kernel the constraints are tighter.
In this case the kernel must be loaded at an offset into system equal
to TEXT_OFFSET - PAGE_OFFSET.
The expected locations for the Device Tree or ATAGs or an initramfs can add more constraints on this LOADADDR.
Putting this in other terms, kernel compilation can generate different images, compressed or not (i.e. needed for XIP).
LOADADDR is the entry_point of the kernel, must be correct (match with where u-boot loads it in DDR) for certain architectures that compiles with absolute long jumps, and not important at all for ARM that can execute relative jumps.
What is the role of ZTEXTADDR in Linux kernel ?
From lxr.linux.no, it's an address in RAM that holds address of zImage as sequence below?
A.
uImage (DataFlash/NAND) ---load_to_RAM--->
uImage (# boot_addr) ---decompress_uImage-->
zImage (# ZTEXTADDR) --- decompress_zImage--->
uncompressed image (# ZRELADDR).
or just:
B.
uImage (DataFlash/NAND) ---load_to_RAM--->
uImage (# boot_addr) ---decompress_uImage-->
uncompressed image (# ZRELADDR)
no using ZTEXTADDR in new kernel version for booting process ?
The Linux ARM decompression boot loader is capable of relocating itself when running from RAM. The relocation portion is PC-relative and so it can be loaded at any address. However, if your main image starts from FLASH/ROM, the code can not be relocated; while moving an image in RAM is a simple memmove(), it is much more involved for NOR flash and can be impossible for ROM.
In this case, a compressed boot linker script is used with the ZTEXTADDR as the location of the decompression code. In your diagram, you have a u-boot, which will load the uImage. There is no reason to execute this directly from Flash/ROM. u-boot can copy the image to RAM and there is no need for the ZTEXTADDR value and it should be left as zero.
If your image boots directly from Flash/ROM, without a boot loader then ZTEXTADDR is useful,
zImage (in flash) --> decompress vmlinux.bin to RAM --> run kernel
The zImage may need to be annotated with some chip setup for this to work and would need ATAGS or device trees linked. For this reason, there are many machine variants in boot/compressed; This is not maintainable and these type of files are discouraged. Typically another bootloader loads the image to RAM and the zImage can move itself to whatever destination it needs; I think that is your situation and you should set ZTEXTADDR to zero and forget about it.
I've been looking into Window's PE format lately and I have noticed that in most examples,
people tend to set the ImageBase offset value in the optional header to something unreasonably high like 0x400000.
What could make it unfavorable not to map an image at offset 0x0?
First off, that's not a default of Windows or the PE file format, it is the default for the linker's /BASE option when you use it to link an EXE. The default for a DLL is 0x10000000.
Selecting /BASE:0 would be bad choice, no program can ever run at that base address. The first 64 KB of the address space is reserved and can never be mapped. Primarily to catch null pointer dereference bugs. And expanded to 64KB to catch pointer bugs in programs that started life in 16-bits and got recompiled to 32-bits.
Why 0x40000 and not 0x10000 is the default is a historical accident as well and goes back to at least Windows 95. Which reserved the first 4 megabytes of the address space for the "16-bit/MS-DOS Compatibility Arena". I don't remember much about it, Windows 9x had a very different 16-bit VM implementation from NT. You can read some more about it in this ancient KB article. It certainly isn't relevant anymore these days, a 64-bit OS will readily allocate heap memory in the space between 0x010000 and 0x400000.
There isn't any point in changing the /BASE option for an EXE. However, there's lots of point changing it for a DLL. They are much more effective if they don't overlap and thus don't have to be relocated, they won't take any space in the paging file and can be shared between processes. There's even an SDK tool for it so you can change it after building, rebase.exe
Practically, the impact of setting /BASE to 0 depends on the Address Space Layout Randomization (ASLR) setting of your image (which is also put by the Linker - /DYNAMICBASE:NO).
Should your image have /BASE:0 and ASLR is on (/DYNAMICBASE:YES), then your image will start and run because the loader will automatically load it at a "valid" address.
Should your image have /BASE:0 and ASLR is off (/DYNAMICBASE:NO), then your image will NOT start because the loader will NOT load it at the desired based address (which is, as explained above, unvalid/reserved).
If you map it to address 0 then that means the code expects to be running starting at address zero.
For the OS, address zero is NULL, which is an invalid address.
(Not "fundamentally", but for modern-day OSes, it is.)
Also, in general you don't want anything in the lower 16 MiB of memory (even virtual), for numerous reasons.
But what's the alternative? It has to be mapped somewhere, so they chose 0x400000... no particular reason for that particular address, probably. It was probably just handy.
Microsoft chose that address as the default starting address specified by the linker at which the PE file will be memory mapped. The linker assumes this address and and can optimize the executable with that assumption. When the file is memory mapped at that address the code can be run without needing to modify any internal offsets.
If for some reason the file cannot be loaded to that location (another exe/dll already loaded there) relocations will need to occur before the executable can run which will increase load times.
Lower memory addresses are usually assumed to contain low level system routines and are generally left alone. The only real requirement for the ImageBase address is that it is a multiple of 0x10000.
Recommended reading:
http://msdn.microsoft.com/en-us/library/ms809762.aspx
From my understanding, after a PC/embedded system booted up, the OS will occupy the entire RAM region, the RAM will look like this:
Which means, while I'm running a program I write, all the variables, dynamic memory allocated in the stacks, heaps and etc, will remain inside the region. If I run firefox, paint, gedit, etc, they will also be running in this region. (Is this understanding correct?)
However, I would like to shrink the OS region. Below is an illustration of how I want to divide the RAM:
The reason that I want to do this is because, I want to store some data receive externally through the driver into the Custom Region at fixed physical location, then I will be able to access it directly from the user space without using copy_to_user().
I think it is possible to do that by configuring u-boot, but I have no experience in u-boot, can anyone give me some directions where to begin with, such as: do I need to modify the source of u-boot, or changing the environment variables of u-boot will be sufficient?
Or is there any alternative method of doing this?
Any help is much appreciated. Thanks!
p/s: I'm using TI ARM processor, and booting up from an SD card, I'm not sure if it matters.
The platform is ARM. min_addr and max_addr will not work on these platform since these are for Intel-only implementations.
For the ARM platform try to look at "mem=size#start" kernel parameter. Read up on Documentation/kernel-parameters.txt and arch/arm/kernel/setup.c. This option is available on most new Linux code base (ie. 2.6.XX).
You need to set the following parameters:
max_addr=some_max_physical
min_addr=some_min_physical
to be passed to the kernel through uboot in the 'bootargs' u-boot environment variable.
I found myself trying to do the opposite recently - in other words get Linux to use the additional memory in my system - although I'm using Barebox rather than u-boot on a OMAP4 platform.
I found (a bit to my surprise) that once the Barebox MLO first stage boot-loader was aware of the extra RAM, the kernel then detected and used it as well without any bootargs. Since the memory size is not passed anywhere on the boot-line, I can only assume the kernel inspects the memory mappings set up by the boot-loader to determine RAM size. This suggests that modifying your u-boot to not map all of the RAM is the way to go.
On the subject of boot-args, there was a time when you it was recommended that you mapped out a chunk of RAM (used by the frame buffer?) on OMAP4 systems, using the boot-line. It's still unclear whether this is still necessary.
I'm going to write and test a bootloader. In order to do this, I am planning to copy the bootloader onto a floppy image file and mount it in a VM.
However, I'm not sure where to put the bootloader's machine code. Does it just get dumped into the first few bytes of the file?
The boot sector of the floppy was the first sector. If you're talking about a raw floppy image (1440K), it should be the first 512 bytes of the image file.
From memory, this gets loaded by the BIOS into 7c00:0000 (real mode) and then jumps to that address.
The DOS boot floppies had a 3-byte JMP instruction there to jump over the Disk Parameter Block (DPB), which detailed the attributes of the disk. But, if you're in total control of the disk and your boot code, I don't think you need to follow that convention. I don't recall any BIOS' checking what was loaded for validity (though admittedly it was a long time ago).
its been a VERY long time but if i recall in DOS it was stored in the MBR. i believe its still the same today
http://en.wikipedia.org/wiki/Master_boot_record